On a Matrix Representation Lemma Useful in Determining Maximal Invariance Groups

Banken (1986, J. Multivariate Anal.19, 156–161) proposed a useful method for determining the group of all affine transformations leaving a multivariate normal testing problem invariant. His main result concerning the derivation of the maximal invariance group is heavily based on a matrix representation lemma which can be considered interesting in its own right. Unfortunately, the proof of this lemma is erroneous and there seems to be no trivial way to correct it. The aim of this note is to show the validity of the assertion.